广义嵌套空间模型变量选择研究——基于广义空间信息准则

发布时间：2018-04-12 21:20

本文选题：空间计量 + 变量选择　；参考：《统计研究》2017年09期

【摘要】：本文提出了广义空间信息准则,以解决广义嵌套空间模型的变量选择问题。依据大样本性质的不同,将该准则分为两类:空间AIC类准则和空间BIC类准则。研究发现,空间AIC类准则能有效解决空间模型中变量的错选和漏选问题,但存在多选变量的倾向;而空间BIC类准则能同时解决空间模型中变量的错选、漏选和多选问题,而且在特殊条件下能更有效解决错选和漏选问题,但往往需要更大的样本容量。Monte Carlo模拟结果印证了上述相关结论。最后,本文以城市对外资银行的吸引力为例,在给定测度指标的基础上,验证其空间相关性,并利用本文提出的方法对其影响因素进行变量选择。
[Abstract]:In this paper, generalized spatial information criterion is proposed to solve the variable selection problem of generalized nested space model.According to the properties of large samples, the criteria are divided into two categories: spatial AIC class criteria and spatial BIC class criteria.It is found that spatial AIC class criterion can effectively solve the problem of misselection and omission of variables in spatial model, but there is a tendency of multi-selection of variables, while the criterion of spatial BIC class can solve the problem of misselection, omission and multi-selection of variables in spatial model at the same time.Moreover, it is more effective to solve the problem of miselection and omission under special conditions, but the simulation results of larger sample size. Monte Carlo often confirm the above conclusions.Finally, taking the attractiveness of cities to foreign banks as an example, on the basis of a given measurement index, this paper verifies its spatial correlation, and selects the variables of its influencing factors by using the method proposed in this paper.
【作者单位】：上海对外经贸大学国际经贸学院;上海财经大学财经研究所;
【基金】：教育部人文社科青年基金“变系数空间面板数据模型变量选择方法及应用”(15YJC790150) 国家统计局全国统计科学研究重点项目“时空经济系统半参数建模方法及其应用”(2016LZ16) 国家自然科学基金面上项目“我国环保产业R&D投入的决策理论与评价方法研究”(71673189)资助
【分类号】：F831.2;O212

【相似文献】